Proyecto final¶

Predicción en el diagnóstico de diabetes¶

Fuente de datos: https://www.kaggle.com/datasets/akshaydattatraykhare/diabetes-dataset

In [5]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.impute import KNNImputer
from sklearn.model_selection import cross_val_score
from sklearn import preprocessing
from sklearn.preprocessing import StandardScaler, RobustScaler
from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn import metrics

Importacion de datos

In [6]:
diabetes = pd.read_csv("diabetes.csv")
diabetes.head()
Out[6]:
Pregnancies Glucose BloodPressure SkinThickness Insulin BMI DiabetesPedigreeFunction Age Outcome
0 6 148 72 35 0 33.6 0.627 50 1
1 1 85 66 29 0 26.6 0.351 31 0
2 8 183 64 0 0 23.3 0.672 32 1
3 1 89 66 23 94 28.1 0.167 21 0
4 0 137 40 35 168 43.1 2.288 33 1

Diccionario:

  • Pregnancies: Cantidad de embarazos por persona
  • Glucose: Nivel de glucosa en la sangre
  • BloodPressure: Medición de presión sanguínea
  • SkinThickness: Grosor de la piel
  • Insulin: nivel de insulina en la sangre
  • BMI: Índice de masa corporal
  • DiabetesPedigreeFunction: Probabilidad de padecer diabeter por historial familiar
  • Age: Edad
  • Outcome: Resultado.
    • 1: tiene diabetes
    • 0: no tiene diabetes

Forma del dataframe

In [7]:
diabetes.shape
Out[7]:
(768, 9)

Todas las columnas tienen tipos acorde a sus valores

In [8]:
diabetes.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 768 entries, 0 to 767
Data columns (total 9 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   Pregnancies               768 non-null    int64  
 1   Glucose                   768 non-null    int64  
 2   BloodPressure             768 non-null    int64  
 3   SkinThickness             768 non-null    int64  
 4   Insulin                   768 non-null    int64  
 5   BMI                       768 non-null    float64
 6   DiabetesPedigreeFunction  768 non-null    float64
 7   Age                       768 non-null    int64  
 8   Outcome                   768 non-null    int64  
dtypes: float64(2), int64(7)
memory usage: 54.1 KB
In [9]:
diabetes.describe()
Out[9]:
Pregnancies Glucose BloodPressure SkinThickness Insulin BMI DiabetesPedigreeFunction Age Outcome
count 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000
mean 3.845052 120.894531 69.105469 20.536458 79.799479 31.992578 0.471876 33.240885 0.348958
std 3.369578 31.972618 19.355807 15.952218 115.244002 7.884160 0.331329 11.760232 0.476951
min 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.078000 21.000000 0.000000
25% 1.000000 99.000000 62.000000 0.000000 0.000000 27.300000 0.243750 24.000000 0.000000
50% 3.000000 117.000000 72.000000 23.000000 30.500000 32.000000 0.372500 29.000000 0.000000
75% 6.000000 140.250000 80.000000 32.000000 127.250000 36.600000 0.626250 41.000000 1.000000
max 17.000000 199.000000 122.000000 99.000000 846.000000 67.100000 2.420000 81.000000 1.000000

Se observa que existen mínimos de 0 para el nivel de glucosa, presión arterial, grosor de la piel, nivel de insulina e índice de masa corporal. Se analizará esta información más adelante.

Se revisan si existen NA en el dataset

In [10]:
diabetes.isna().sum()
Out[10]:
Pregnancies                 0
Glucose                     0
BloodPressure               0
SkinThickness               0
Insulin                     0
BMI                         0
DiabetesPedigreeFunction    0
Age                         0
Outcome                     0
dtype: int64
In [11]:
diabetes.hist(grid=False, bins=15,figsize=(15,12), rwidth=0.85, color="#8470FF", ec="blueviolet")
Out[11]:
array([[<Axes: title={'center': 'Pregnancies'}>,
        <Axes: title={'center': 'Glucose'}>,
        <Axes: title={'center': 'BloodPressure'}>],
       [<Axes: title={'center': 'SkinThickness'}>,
        <Axes: title={'center': 'Insulin'}>,
        <Axes: title={'center': 'BMI'}>],
       [<Axes: title={'center': 'DiabetesPedigreeFunction'}>,
        <Axes: title={'center': 'Age'}>,
        <Axes: title={'center': 'Outcome'}>]], dtype=object)
In [12]:
corr = diabetes.corr()
corr
Out[12]:
Pregnancies Glucose BloodPressure SkinThickness Insulin BMI DiabetesPedigreeFunction Age Outcome
Pregnancies 1.000000 0.129459 0.141282 -0.081672 -0.073535 0.017683 -0.033523 0.544341 0.221898
Glucose 0.129459 1.000000 0.152590 0.057328 0.331357 0.221071 0.137337 0.263514 0.466581
BloodPressure 0.141282 0.152590 1.000000 0.207371 0.088933 0.281805 0.041265 0.239528 0.065068
SkinThickness -0.081672 0.057328 0.207371 1.000000 0.436783 0.392573 0.183928 -0.113970 0.074752
Insulin -0.073535 0.331357 0.088933 0.436783 1.000000 0.197859 0.185071 -0.042163 0.130548
BMI 0.017683 0.221071 0.281805 0.392573 0.197859 1.000000 0.140647 0.036242 0.292695
DiabetesPedigreeFunction -0.033523 0.137337 0.041265 0.183928 0.185071 0.140647 1.000000 0.033561 0.173844
Age 0.544341 0.263514 0.239528 -0.113970 -0.042163 0.036242 0.033561 1.000000 0.238356
Outcome 0.221898 0.466581 0.065068 0.074752 0.130548 0.292695 0.173844 0.238356 1.000000
In [13]:
mask = np.triu(np.ones_like(corr, dtype=bool))
sns.heatmap(corr, mask=mask, linewidths=0.75, annot=True, cmap="BuPu", annot_kws={"size":8})
Out[13]:
<Axes: >
In [14]:
colores = {0: '#3CB371', 1: '#7A378B'}
sns.pairplot(diabetes, hue="Outcome", palette=colores, markers=['o', 'X'])
plt.show()

Nuevamente se observan varios puntos en 0 para el nivel de glucosa, presión arterial, grosor de la piel, nivel de insulina e índice de masa corporal.

Según la Organización Mundial de la Salud, WHO por sus siglas en inglés, los valores normales de glucosa en la sangre se encuentran entre 70mg/dL y 100mg/dL.

Fuente: https://www.who.int/data/gho/indicator-metadata-registry/imr-details/2380

En la gráfica se observa que este comportamiento tambien ocurre para la presión arterial, grosor de la piel, nivel de insulina e índice de masa corporal. Supondremos que la data original contenía valores tipo NAN y fueron reemplazados por 0, ya que no tiene sentido que una persona no tenga glucosa o insulina en la sangre.

Para esto, se imputarán los valores faltantes. Se hará de 2 formas, por media y por KNN.

No se tomará en cuenta la cantidad de embarazos para la imputación ya que sí es coherente que haya mujeres que no han estado embarazadas.

Imputación por media¶

In [15]:
diabetes_Med = diabetes.copy()
columnas = diabetes_Med.columns[1:-1]

for col in columnas:
    diabetes_Med[col] = diabetes_Med[col].replace(0, np.nan)
    media = diabetes_Med[col].mean()
    diabetes_Med[col] = diabetes_Med[col].fillna(media)
In [16]:
colores2 = {0: '#48D1CC', 1: '#7A378B'}
sns.pairplot(diabetes_Med, hue="Outcome", palette=colores2, markers=['o', 'X'])
plt.show()

Imputación por KNN¶

In [17]:
diabetes_KNN = diabetes.copy()
columnas2 = diabetes_KNN.columns[1:-1]

for col in columnas2:
    diabetes_KNN[col] = diabetes_KNN[col].replace(0, np.nan)

diabetes_KNN.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 768 entries, 0 to 767
Data columns (total 9 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   Pregnancies               768 non-null    int64  
 1   Glucose                   763 non-null    float64
 2   BloodPressure             733 non-null    float64
 3   SkinThickness             541 non-null    float64
 4   Insulin                   394 non-null    float64
 5   BMI                       757 non-null    float64
 6   DiabetesPedigreeFunction  768 non-null    float64
 7   Age                       768 non-null    int64  
 8   Outcome                   768 non-null    int64  
dtypes: float64(6), int64(3)
memory usage: 54.1 KB
In [18]:
from scipy.sparse import data
from sklearn.pipeline import Pipeline
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import RepeatedStratifiedKFold

results = list()
strategies = [str(i) for i in [1,3,5,7,9,15,18,21]]
for s in strategies:
 # create the modeling pipeline
 pipeline = Pipeline(steps=[('i', KNNImputer(n_neighbors=int(s))), ('m', RandomForestClassifier())])
 # evaluate the model
 cv = RepeatedStratifiedKFold(n_splits=10, n_repeats=3, random_state=1)
 scores = cross_val_score(pipeline, diabetes_KNN.drop(["Outcome"], axis=1), diabetes_KNN["Outcome"],scoring='accuracy', cv=cv, n_jobs=-1)
 # store results
 results.append(scores)
 print('>%s %.3f (%.3f)' % (s, np.mean(scores), np.std(scores)))

>1 0.752 (0.047)
>3 0.750 (0.048)
>5 0.760 (0.049)
>7 0.749 (0.054)
>9 0.755 (0.052)
>15 0.759 (0.047)
>18 0.758 (0.055)
>21 0.754 (0.053)
In [19]:
imputer = KNNImputer(n_neighbors=15)
diabetes_KNN[columnas2] = imputer.fit_transform(diabetes_KNN[columnas2])
In [20]:
colores3 = {0: '#3A5FCD', 1: '#7A378B'}
sns.pairplot(diabetes_KNN, hue="Outcome", palette=colores3, markers=['o', 'X'])
plt.show()
In [21]:
diabetes_KNN.describe()
Out[21]:
Pregnancies Glucose BloodPressure SkinThickness Insulin BMI DiabetesPedigreeFunction Age Outcome
count 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000 768.000000
mean 3.845052 121.649045 72.331510 29.130208 152.334809 32.439470 0.471876 33.240885 0.348958
std 3.369578 30.459769 12.153261 9.163552 95.001790 6.882721 0.331329 11.760232 0.476951
min 0.000000 44.000000 24.000000 7.000000 14.000000 18.200000 0.078000 21.000000 0.000000
25% 1.000000 99.750000 64.000000 23.000000 89.283333 27.500000 0.243750 24.000000 0.000000
50% 3.000000 117.000000 72.000000 29.166667 138.800000 32.250000 0.372500 29.000000 0.000000
75% 6.000000 140.250000 80.000000 34.033333 185.000000 36.600000 0.626250 41.000000 1.000000
max 17.000000 199.000000 122.000000 99.000000 846.000000 67.100000 2.420000 81.000000 1.000000
Boxplot Diabetes¶
In [22]:
for col in diabetes.columns[0:-1]:
    plt.figure() 
    sns.boxplot(x=col, data=diabetes, orient="v", color="#3CB371", notch=True, linewidth=2, width=.35)

c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
Boxplot Diabetes_Med¶
In [23]:
for col in diabetes_Med.columns[0:-1]:
    plt.figure() 
    sns.boxplot(x=col, data=diabetes_Med, orient="v", color="#48D1CC", notch=True, linewidth=2, width=.35)

c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
Dataset Diabetes_KNN¶
In [24]:
for col in diabetes_KNN.columns[0:-1]:
    plt.figure() 
    sns.boxplot(x=col, data=diabetes_KNN, orient="v", color="#3A5FCD", notch=True, linewidth=2, width=.35)

c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))
c:\Users\and_w\anaconda3\envs\PruebaClase\lib\site-packages\seaborn\_oldcore.py:1599: UserWarning: Vertical orientation ignored with only `x` specified.
  warnings.warn(single_var_warning.format("Vertical", "x"))

Se observa que aún existen outliers luego de la imputación. Se desplegarán los valores más grandes de cada columna con outliers en los datasets imputados. No se manipularán los datos del dataset original.

In [25]:
diabetes_Med.columns
Out[25]:
Index(['Pregnancies', 'Glucose', 'BloodPressure', 'SkinThickness', 'Insulin',
       'BMI', 'DiabetesPedigreeFunction', 'Age', 'Outcome'],
      dtype='object')
In [26]:
print(diabetes_Med["Glucose"].nlargest(10))

661    199.0
561    198.0
8      197.0
228    197.0
408    197.0
579    197.0
22     196.0
206    196.0
359    196.0
498    195.0
Name: Glucose, dtype: float64
In [27]:
print(diabetes_Med["BloodPressure"].nsmallest(10))

597    24.0
18     30.0
125    30.0
599    38.0
4      40.0
80     44.0
575    44.0
576    44.0
741    44.0
346    46.0
Name: BloodPressure, dtype: float64
In [28]:
diabetes_Med.drop(index=[597,18,125], inplace=True)
In [29]:
print(diabetes_Med["SkinThickness"].nlargest(10))

579    99.0
445    63.0
57     60.0
120    56.0
86     54.0
211    54.0
275    52.0
532    52.0
99     51.0
150    50.0
Name: SkinThickness, dtype: float64
In [30]:
diabetes_Med.drop(index=579, inplace=True)
In [31]:
print(diabetes_Med["Insulin"].nlargest(10))

13     846.0
228    744.0
247    680.0
584    600.0
409    579.0
286    545.0
8      543.0
655    540.0
753    510.0
111    495.0
Name: Insulin, dtype: float64
In [32]:
diabetes_Med.drop(index=[13,228,247], inplace=True)
In [33]:
print(diabetes_Med["BMI"].nlargest(15))

177    67.1
445    59.4
673    57.3
120    53.2
303    52.9
193    52.3
155    50.0
99     49.7
681    49.6
746    49.3
84     48.8
378    48.3
154    47.9
335    47.9
57     46.8
Name: BMI, dtype: float64
In [34]:
diabetes_Med.drop(index=177, inplace=True)
In [35]:
diabetes_KNN.columns
Out[35]:
Index(['Pregnancies', 'Glucose', 'BloodPressure', 'SkinThickness', 'Insulin',
       'BMI', 'DiabetesPedigreeFunction', 'Age', 'Outcome'],
      dtype='object')
In [36]:
print(diabetes_KNN["Glucose"].nlargest(10))

661    199.0
561    198.0
8      197.0
228    197.0
408    197.0
579    197.0
22     196.0
206    196.0
359    196.0
498    195.0
Name: Glucose, dtype: float64
In [37]:
print(diabetes_KNN["BloodPressure"].nsmallest(10))

597    24.0
18     30.0
125    30.0
599    38.0
4      40.0
80     44.0
575    44.0
576    44.0
741    44.0
346    46.0
Name: BloodPressure, dtype: float64
In [38]:
diabetes_KNN.drop(index=[597,18,125], inplace=True)
In [39]:
print(diabetes_KNN["SkinThickness"].nlargest(10))

579    99.0
445    63.0
57     60.0
120    56.0
86     54.0
211    54.0
275    52.0
532    52.0
99     51.0
150    50.0
Name: SkinThickness, dtype: float64
In [40]:
diabetes_KNN.drop(index=579, inplace=True)
In [41]:
print(diabetes_KNN["Insulin"].nlargest(10))

13     846.0
228    744.0
247    680.0
584    600.0
409    579.0
286    545.0
8      543.0
655    540.0
753    510.0
111    495.0
Name: Insulin, dtype: float64
In [42]:
diabetes_KNN.drop(index=[13,228,247], inplace=True)
In [43]:
print(diabetes_KNN["BMI"].nlargest(15))

177    67.1
445    59.4
673    57.3
120    53.2
303    52.9
193    52.3
155    50.0
99     49.7
681    49.6
746    49.3
84     48.8
378    48.3
154    47.9
335    47.9
57     46.8
Name: BMI, dtype: float64
In [44]:
diabetes_KNN.drop(index=177, inplace=True)

Se comparan las densidades resultantes luego de la imputación y eliminación de outliers

In [45]:
for col in diabetes.columns[0:-1]:
    fig, ax = plt.subplots()
    sns.kdeplot(diabetes[col], ax=ax, label='Diabetes', fill=True)
    sns.kdeplot(diabetes_Med[col], ax=ax, label='Diabetes Media', fill=True)
    sns.kdeplot(diabetes_KNN[col], ax=ax, label='Diabetes KNN', fill=True)
    ax.set_ylabel('Densidad')
    ax.legend()
    plt.show()

Datasets de entrenamiento y prueba¶

In [46]:
# diabetes
x = diabetes.drop(["Outcome"], axis = 1)
y = diabetes["Outcome"]

# diabetes media imputada
x_med = diabetes_Med.drop(["Outcome"], axis = 1)
y_med = diabetes_Med["Outcome"]

# diabetes KNN imputado
x_KNN = diabetes_KNN.drop(["Outcome"], axis = 1)
y_KNN = diabetes_KNN["Outcome"]
In [47]:
seed = 42

# diabetes
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size = 0.3, random_state = seed)

# diabetes media imputada
xmed_train, xmed_test, ymed_train, ymed_test = train_test_split(x_med, y_med, test_size = 0.3, random_state = seed)

# diabetes KNN imputado
xknn_train, xknn_test, yknn_train, yknn_test = train_test_split(x_KNN, y_KNN, test_size = 0.3, random_state = seed)

Estandarización de datos¶

Se utilizarán dos métodos para la estandarización. La estándar y la robusta. La estandarización simple transforma los datos según la media y la desviación estándar, mientras que la robusta utiliza la mediana y el rango intercuantílico.

In [48]:
estandar = StandardScaler()
robusta = RobustScaler()

#Estándar
sx_train = estandar.fit_transform(x_train)
sx_test = estandar.transform(x_test)

sxmed_train = estandar.fit_transform(xmed_train)
sxmed_test = estandar.transform(xmed_test)

sxknn_train = estandar.fit_transform(xknn_train)
sxknn_test = estandar.transform(xknn_test)

#Robusta
rsx_train = robusta.fit_transform(x_train)
rsx_test = robusta.transform(x_test)

rsxmed_train = robusta.fit_transform(xmed_train)
rsxmed_test = robusta.transform(xmed_test)

rsxknn_train = robusta.fit_transform(xknn_train)
rsxknn_test = robusta.transform(xknn_test)

Cross Validation para encontrar el mejor K¶

Para el dataset original con estandarización simple

In [49]:
rango = [i for i in range (1,11)]
acc = []

for k in rango:
    knn = KNeighborsClassifier(n_neighbors=k)
    score = cross_val_score(knn, sx_train, y_train, cv=5)
    acc.append(np.mean(score))
In [50]:
sns.lineplot(x = rango, y = acc, marker = 'o')
plt.xlabel("Valor de K")
plt.ylabel("Accuracy")
Out[50]:
Text(0, 0.5, 'Accuracy')

Para el dataset imputado con la media y estandarización simple

In [77]:
rango2 = [i for i in range (1,31)]
acc2 = []

for k in rango2:
    knn = KNeighborsClassifier(n_neighbors=k)
    score = cross_val_score(knn, sxmed_train, ymed_train, cv=5)
    acc2.append(np.mean(score))
In [78]:
sns.lineplot(x = rango2, y = acc2, marker = 'o')
plt.xlabel("Valor de K")
plt.ylabel("Accuracy")
Out[78]:
Text(0, 0.5, 'Accuracy')

Para el dataset imputado con KNN y con estandarización simple

In [53]:
rango3 = [i for i in range (1,31)]
acc3 = []

for k in rango3:
    knn = KNeighborsClassifier(n_neighbors=k)
    score = cross_val_score(knn, sxknn_train, yknn_train, cv=5)
    acc3.append(np.mean(score))
In [54]:
sns.lineplot(x = rango3, y = acc3, marker = 'o')
plt.xlabel("Valor de K")
plt.ylabel("Accuracy")
Out[54]:
Text(0, 0.5, 'Accuracy')

Para el dataset original con estandarización robusta

In [55]:
rango4 = [i for i in range (1,21)]
acc4 = []

for k in rango4:
    knn = KNeighborsClassifier(n_neighbors=k)
    score = cross_val_score(knn, rsx_train, y_train, cv=5)
    acc4.append(np.mean(score))
In [56]:
sns.lineplot(x = rango4, y = acc4, marker = 'o')
plt.xlabel("Valor de K")
plt.ylabel("Accuracy")
Out[56]:
Text(0, 0.5, 'Accuracy')

Para el dataset imputado con la media y estandarización robusta

In [57]:
rango5 = [i for i in range (1,31)]
acc5 = []

for k in rango5:
    knn = KNeighborsClassifier(n_neighbors=k)
    score = cross_val_score(knn, rsxmed_train, ymed_train, cv=5)
    acc5.append(np.mean(score))
In [58]:
sns.lineplot(x = rango5, y = acc5, marker = 'o')
plt.xlabel("Valor de K")
plt.ylabel("Accuracy")
Out[58]:
Text(0, 0.5, 'Accuracy')

Para el dataset imputado con KNN y con estandarización robusta

In [59]:
rango6 = [i for i in range (1,31)]
acc6 = []

for k in rango6:
    knn = KNeighborsClassifier(n_neighbors=k)
    score = cross_val_score(knn, rsxknn_train, yknn_train, cv=5)
    acc6.append(np.mean(score))
In [60]:
sns.lineplot(x = rango6, y = acc6, marker = 'o')
plt.xlabel("Valor de K")
plt.ylabel("Accuracy")
Out[60]:
Text(0, 0.5, 'Accuracy')
In [61]:
mejor_k = rango[np.argmax(acc)]
mejor_kmed = rango2[np.argmax(acc2)]
mejor_kknn = rango3[np.argmax(acc3)]
mejor_rk = rango4[np.argmax(acc4)]
mejor_rkmed = rango5[np.argmax(acc5)]
mejor_rkknn = rango6[np.argmax(acc6)]

Entrenamiento de los modelos¶

In [62]:
# Estandarización simple
knn = KNeighborsClassifier(n_neighbors=mejor_k)
knn.fit(sx_train, y_train)

knn_med = KNeighborsClassifier(n_neighbors=mejor_kmed)
knn_med.fit(sxmed_train, ymed_train)

knn_knn = KNeighborsClassifier(n_neighbors=mejor_kknn)
knn_knn.fit(sxknn_train, yknn_train)

# Estandarización robusta
rknn = KNeighborsClassifier(n_neighbors=mejor_rk)
rknn.fit(rsx_train, y_train)

rknn_med = KNeighborsClassifier(n_neighbors=mejor_rkmed)
rknn_med.fit(rsxmed_train, ymed_train)

rknn_knn = KNeighborsClassifier(n_neighbors=mejor_rkknn)
rknn_knn.fit(rsxknn_train, yknn_train)
Out[62]:
KNeighborsClassifier(n_neighbors=27)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
KNeighborsClassifier(n_neighbors=27)

Predicción de los modelos¶

In [63]:
y_pred = knn.predict(sx_test)
ymed_pred = knn_med.predict(sxmed_test)
yknn_pred = knn_knn.predict(sxknn_test)
In [79]:
yr_pred = rknn.predict(rsx_test)
yrmed_pred = rknn_med.predict(rsxmed_test)
yrknn_pred = rknn_knn.predict(rsxknn_test)

Evaluación de los modelos¶

In [64]:
print(classification_report(y_test, y_pred))

              precision    recall  f1-score   support

           0       0.75      0.80      0.78       151
           1       0.57      0.50      0.53        80

    accuracy                           0.70       231
   macro avg       0.66      0.65      0.65       231
weighted avg       0.69      0.70      0.69       231
In [65]:
print(classification_report(ymed_test, ymed_pred))

              precision    recall  f1-score   support

           0       0.82      0.81      0.82       156
           1       0.60      0.61      0.61        72

    accuracy                           0.75       228
   macro avg       0.71      0.71      0.71       228
weighted avg       0.75      0.75      0.75       228
In [66]:
print(classification_report(yknn_test, yknn_pred))

              precision    recall  f1-score   support

           0       0.80      0.84      0.82       156
           1       0.61      0.54      0.57        72

    accuracy                           0.75       228
   macro avg       0.70      0.69      0.70       228
weighted avg       0.74      0.75      0.74       228
In [67]:
print(classification_report(y_test, yr_pred))

              precision    recall  f1-score   support

           0       0.76      0.85      0.81       151
           1       0.65      0.50      0.56        80

    accuracy                           0.73       231
   macro avg       0.70      0.68      0.68       231
weighted avg       0.72      0.73      0.72       231
In [68]:
print(classification_report(ymed_test, yrmed_pred))

              precision    recall  f1-score   support

           0       0.81      0.82      0.81       156
           1       0.59      0.57      0.58        72

    accuracy                           0.74       228
   macro avg       0.70      0.69      0.70       228
weighted avg       0.74      0.74      0.74       228
In [69]:
print(classification_report(yknn_test, yrknn_pred))

              precision    recall  f1-score   support

           0       0.78      0.85      0.81       156
           1       0.59      0.49      0.53        72

    accuracy                           0.73       228
   macro avg       0.69      0.67      0.67       228
weighted avg       0.72      0.73      0.72       228

De los modelos considerados para KNN, el que obtuvo mejor resultado fue el modelo utilizando el dataset con media imputada y estandarización simple. Sin embargo, la exactitud y precisión al momento de detectar casos con diabetes no es la mejor. Por lo que se evaluará un modelo diferente.

Árbol de decisiones¶

In [70]:
from sklearn.tree import DecisionTreeClassifier
In [71]:
arbol = DecisionTreeClassifier(criterion="entropy", max_depth= 3)
arbol_med = DecisionTreeClassifier(criterion="entropy", max_depth = 3)
arbol_knn = DecisionTreeClassifier(criterion="entropy", max_depth = 3)
In [72]:
arbol.fit(sx_train, y_train)
arbol_med.fit(sxmed_train, ymed_train)
arbol_knn.fit(sxknn_train, yknn_train)
Out[72]:
DecisionTreeClassifier(criterion='entropy', max_depth=3)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
DecisionTreeClassifier(criterion='entropy', max_depth=3)
In [73]:
arbol_pred = arbol.predict(sx_test)
arbol_predmed = arbol_med.predict(sxmed_test)
arbol_predknn = arbol_knn.predict(sxknn_test)
In [74]:
print(classification_report(y_test, arbol_pred))

              precision    recall  f1-score   support

           0       0.72      0.93      0.81       151
           1       0.71      0.31      0.43        80

    accuracy                           0.72       231
   macro avg       0.72      0.62      0.62       231
weighted avg       0.72      0.72      0.68       231
In [75]:
print(classification_report(ymed_test, arbol_predmed))

              precision    recall  f1-score   support

           0       0.81      0.84      0.83       156
           1       0.63      0.58      0.60        72

    accuracy                           0.76       228
   macro avg       0.72      0.71      0.72       228
weighted avg       0.75      0.76      0.76       228
In [76]:
print(classification_report(yknn_test, arbol_predknn))

              precision    recall  f1-score   support

           0       0.75      0.96      0.84       156
           1       0.77      0.32      0.45        72

    accuracy                           0.75       228
   macro avg       0.76      0.64      0.65       228
weighted avg       0.76      0.75      0.72       228

Utilizando un modelo de árbol de decisiones, el dataset con la imputación por media sigue siendo el mejor de todos los modelos.